Messenger RNA precursors are often processed through alternative splicing into multiple mRNA isoforms, which in many cases encode proteins with distinct functions. It is estimated that as many as 30 percent of genes in humans exhibit alternative splicing and many alternative splicing events are known to be associated with human diseases. However, the current technology to analyze alternative splicing is slow and labor-intensive. Here we propose to develop a novel system to detect alternative splicing on a large scale and apply the technology to the molecular classification of cancer. Our experimental strategy is to synthesize DNA oligos that are selectively hybridized to spliced mRNAs through specific splice junction sequences. Each oligo will be linked to a unique 20 nt sequence to serve as an addressable "zip-code". Pooled oligos will be hybridized to total RNA from cells or tissues, and those annealed to mRNAs will be isolated by poly (A)+ selection followed by PCR-amplification. The products will then be hybridized to a universal zip-code array manufactured by Illumina based in San Diego. Individual splicing events will then be detected and quantified by imaging the array. This technology is different from any existing methodology in several important aspects: (1) Our experiments are designed to detect mRNA isoforms, rather than overall gene expression. This will allow us to specifically relate the post-transcriptional processing step to important biological processes such as cancer. (2) Our procedure, is highly scaleable, which will allow us to approach the alternative splicing problem at the genomic level. (3) Through a collaborative arrangement, we will have access to state-of-the-art array technology developed at Illumina, which is designed for large scale analyses in a "array of arrays" format. (4) Our strategy allows customization of the assay for specific applications and provides maximum flexibility in experimental design with minimal costs. We plan to approach the alternative splicing problem in a systematic way from database construction to assay development to application of the technology in cancer classification. For this purpose, we have assembled a team consisting of leading experts with highly complementary expertise. Because many alternative splicing events have been individually linked to specific cancer or cancer stages, it is almost certain that this combined approach will lead to efficient and accurate predictors for various diseases, which will be highly complementary to those based on monitoring gene expression. In fact, because alternative splicing is a general biological process, we believe that our technology will provide an urgently needed tool for both basic and clinical research, and therefore, will have a broad impact in many fields.